create-model-yml.yml

name: Create HuggingFace Model Repository

on:
  workflow_dispatch:
    inputs:
      model_name:
        description: "Name of the model to create (will be used in repo name and files)"
        required: true
        type: string
      model_variant:
        description: "Model variant (e.g., 8b, 13b) - will be used in default field"
        required: true
        type: string
        default: "8b"
      prompt_template:
        description: "Prompt template for the model"
        type: string
        default: |
          <|im_start|>system
          {system_message}<|im_end|>
          <|im_start|>user
          {prompt}<|im_end|>
          <|im_start|>assistant
      stop_tokens:
        description: "Stop tokens for the model (comma-separated, e.g., <|im_end|>,</s>)"
        type: string
        default: "<|im_end|>"
      engine:
        description: "Engine to run the model (e.g., llama-cpp)"
        required: true
        type: string
        default: "llama-cpp"

env:
  USER_NAME: cortexso
  MODEL_NAME: ${{ inputs.model_name }}
  MODEL_VARIANT: ${{ inputs.model_variant }}
  PROMPT_TEMPLATE: ${{ inputs.prompt_template }}
  STOP_TOKENS: ${{ inputs.stop_tokens }}
  ENGINE: ${{ inputs.engine }}
  
jobs:
  create-repo:
    runs-on: ubuntu-20-04-gguf
    timeout-minutes: 7200
    steps:
      - name: Checkout
        uses: actions/checkout@v4

      - name: Set up Python
        uses: actions/setup-python@v5
        with:
          python-version: '3.10'
      
      - name: Cache Python packages
        uses: actions/cache@v4
        with:
          path: |
            ~/.cache/pip
            ~/.local/share/pip
            .venv
          key: ${{ runner.os }}-pip-${{ github.sha }}
          restore-keys: |
            ${{ runner.os }}-pip-

      - name: Install dependencies
        run: |
          pip install huggingface_hub PyYAML
          git lfs install

      - name: Create YAML files
        run: |
          python3 - << EOF
          import yaml
          import os
          from yaml.representer import SafeRepresenter

          class CustomDumper(yaml.SafeDumper):
              pass

          def custom_str_representer(dumper, data):
              if isinstance(data, str) and '\n' in data:
                  return dumper.represent_scalar('tag:yaml.org,2002:str', data, style='|')
              return SafeRepresenter.represent_str(dumper, data)

          CustomDumper.add_representer(str, custom_str_representer)

          def dict_representer(dumper, data):
              return dumper.represent_dict(data.items())

          CustomDumper.add_representer(dict, dict_representer)

          # Process stop tokens
          stop_tokens = os.environ['STOP_TOKENS'].split(',')
          stop_tokens = [token.strip() for token in stop_tokens if token.strip()]

          # Create model.yml content with sections
          def write_section(file, content, section_name='', subsection=''):
              if section_name:
                  file.write(f"# BEGIN {section_name}\n")
              if subsection:
                  file.write(f"# BEGIN {subsection}\n")
              
              yaml_content = yaml.dump(content, Dumper=CustomDumper, default_flow_style=False, sort_keys=False)
              file.write(yaml_content)
                        
              if subsection:
                  file.write(f"# END {subsection}\n")
              if section_name:
                  file.write(f"# END {section_name}\n")

          with open('model.yml', 'w') as f:
              # General metadata section
              general_metadata = {
                  'id': os.environ['MODEL_NAME'],
                  'model': os.environ['MODEL_NAME'],
                  'name': os.environ['MODEL_NAME'],
                  'version': 1
              }
              write_section(f, general_metadata, 'GENERAL GGUF METADATA')
              f.write('\n')

              # Inference parameters section
              f.write("# BEGIN INFERENCE PARAMETERS\n")
              
              # Required subsection
              required_inference = {'stop': stop_tokens}
              write_section(f, required_inference, '', 'REQUIRED')
              f.write('\n')
              
              # Optional subsection
              optional_inference = {
                  'stream': True,
                  'top_p': 0.9,
                  'temperature': 0.7,
                  'frequency_penalty': 0,
                  'presence_penalty': 0,
                  'max_tokens': 4096,
                  'seed': -1,
                  'dynatemp_range': 0,
                  'dynatemp_exponent': 1,
                  'top_k': 40,
                  'min_p': 0.05,
                  'tfs_z': 1,
                  'typ_p': 1,
                  'repeat_last_n': 64,
                  'repeat_penalty': 1,
                  'mirostat': False,
                  'mirostat_tau': 5,
                  'mirostat_eta': 0.100000001,
                  'penalize_nl': False,
                  'ignore_eos': False,
                  'n_probs': 0,
                  'min_keep': 0
              }
              write_section(f, optional_inference, '', 'OPTIONAL')
              f.write("# END INFERENCE PARAMETERS\n\n")

              # Model load parameters section
              f.write("# BEGIN MODEL LOAD PARAMETERS\n")
              required_load = {
                  'engine': os.environ['ENGINE'],
                  'prompt_template': os.environ['PROMPT_TEMPLATE'],
                  'ctx_len': 4096,
                  'ngl': 34
              }
              write_section(f, required_load, '', 'REQUIRED')
              f.write("# END MODEL LOAD PARAMETERS\n")

          # Create metadata.yml with dynamic default field
          metadata_content = {
              'version': 1,
              'name': os.environ['MODEL_NAME'],
              'default': f"{os.environ['MODEL_VARIANT']}-gguf-q4-km"
          }

          with open('metadata.yml', 'w') as f:
              write_section(f, metadata_content) 
          EOF

      - name: Create HuggingFace Repository and Upload Files
        env:
          HF_TOKEN: ${{ secrets.HUGGINGFACE_TOKEN_WRITE }}
        run: |
          python3 - << EOF
          from huggingface_hub import HfApi, create_repo
          import os

          # Initialize the Hugging Face API
          api = HfApi(token=os.environ['HF_TOKEN'])

          # Create the repository
          repo_id = f"${{ env.USER_NAME }}/${{ env.MODEL_NAME }}"
          try:
              create_repo(repo_id, private=False, token=os.environ['HF_TOKEN'])
              print(f"Created repository: {repo_id}")
          except Exception as e:
              print(f"Repository might already exist or error occurred: {e}")

          # Upload the files
          api.upload_file(
              path_or_fileobj="model.yml",
              path_in_repo="model.yml",
              repo_id=repo_id,
              token=os.environ['HF_TOKEN']
          )
          api.upload_file(
              path_or_fileobj="metadata.yml",
              path_in_repo="metadata.yml",
              repo_id=repo_id,
              token=os.environ['HF_TOKEN']
          )
          print("Files uploaded successfully")
          EOF

      - name: Cleanup
        run: |
          rm -f model.yml metadata.yml